Infrastructure firms face huge losses if AI fails to replace 250 million jobs
An analyst says the current spending on AI hardware requires massive workforce reduction to justify the costs, raising fears of a bubble.
Key points
- AI must generate $2.5 trillion in annual revenue to justify current investments.
- Current AI revenues are estimated at only $150 billion to $250 billion yearly.
- Infrastructure providers face high depreciation costs as hardware becomes obsolete quickly.
An analyst warns that the massive spending on artificial intelligence infrastructure will only be justified if the technology replaces 250 million jobs, a number that exceeds the total nonfarm workforce in the United States. This discrepancy suggests that infrastructure providers could face severe financial losses if AI does not deliver the expected returns on investment.
In plain English
The core issue is a mismatch between spending and earnings. Companies have poured trillions of dollars into building the physical systems that run AI, such as data centers and power grids. To make this money back, the AI services must generate enormous amounts of revenue. According to Marin Ivezic, founder of Applied Quantum, the math requires AI to produce $2.5 trillion in annual income. Current estimates place AI revenue far below this target, creating a significant financial gap that threatens the stability of the providers who built the infrastructure.
The background
Since late 2022, developers have engaged in a rapid expansion of hardware and energy capacity to support AI growth. Ivezic notes that $1.5 trillion has already been spent on hardware since then, with an additional $1 trillion expected to be invested annually. A major complication is the rapid obsolescence of high-performance AI chips, which become outdated roughly every five years. In contrast, the data centers housing these chips are designed to last about 20 years. This mismatch creates heavy depreciation costs for infrastructure owners.
What changes now
Ivezic describes a situation he calls the “50-Cent Rule,” where providers must earn about 50 cents in annual revenue for every dollar invested to break even. The challenge is compounded by the fact that AI models are becoming cheaper and more capable. For instance, models with GPT-4-class capabilities have dropped in price by a factor of 100 since 2022. While end-users pay less, the costs for infrastructure providers have not decreased and are actually rising. This dynamic makes it increasingly difficult for these firms to recover their initial investments, echoing concerns about a potential bubble similar to the dot-com era.
Key terms explained
Infrastructure providers are the companies that build and maintain the physical hardware, such as servers and data centers, required to run AI models. Depreciation refers to the loss of value of these assets over time, especially when technology becomes obsolete quickly. Return on investment is the financial gain made from spending money on these systems. Teams usually monitor these metrics to ensure that the cost of maintaining hardware does not exceed the revenue generated by the services running on it.
What to do and how to stay safe: Applied Quantum
- Review current contracts with infrastructure providers to understand how hardware refresh cycles and depreciation costs are handled.
- Monitor pricing trends for AI services to anticipate shifts in operating costs as models become more efficient and cheaper.
- Evaluate the long-term viability of AI projects by comparing projected revenue against the substantial capital expenditure required for hardware.
- Watch for market signals indicating financial strain among data center operators, which could affect service stability or pricing.
Step-by-step guide: Stop AI Chatbots Leaking Sensitive Business Data
General security guidance from the Vector Update newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How much revenue is needed to justify current AI investments?
According to Marin Ivezic, AI needs to generate approximately $2.5 trillion in annual revenue to justify the current investment levels.
What is the current estimated revenue of the AI industry?
Optimistic estimates place current AI revenues between $150 billion and $250 billion per year, which is far below the required amount.
Why is it hard for infrastructure providers to make a profit?
Providers face high depreciation costs because AI chips become obsolete every five years, while data centers last 20 years, and hardware costs are rising while service prices drop.
